Mitra: A Scalable Continuous Media Server

  • Authors:

  • Shahram Ghandeharizadeh;Roger Zimmermann;Weifeng Shi;Reza Rejaie;Doug Ierardi;Ta-Wei Li

  • Affiliations:

  • Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu;Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu;Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu;Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu;Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu;Computer Science Department, University of Southern California, Los Angeles, California 90089. E-mail: shahram@research.panasonic.com/ ierardi@cs.usc.edu

  • Venue:
  • Multimedia Tools and Applications
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Mitra is a scalable storage manager that supports the display ofcontinuous media data types, e.g., audio and video clips. It is asoftware based system that employs off-the-shelf hardware components.Its present hardware platform is a cluster of multi-diskworkstations, connected using an ATM switch. Mitra supports thedisplay of a mix of media types. To reduce the cost of storage, itsupports a hierarchical organization of storage devices and stagesthe frequently accessed objects on the magnetic disks. For the numberof displays to scale as a function of additional disks, Mitra employsstaggered striping. It implements three strategies to maximize thenumber of simultaneous displays supported by each disk. First, theEVEREST file system allows different files (corresponding to objectsof different media types) to be retrieved at different block sizegranularities. Second, the FIXB algorithm recognizes the differentzones of a disk and guarantees a continuous display while harnessingthe average disk transfer rate. Third, Mitra implements the GroupedSweeping Scheme (GSS) to minimize the impact of disk seeks on theavailable disk bandwidth.In addition to reporting on implementation details of Mitra, wepresent performance results that demonstrate the scalabilitycharacteristics of the system. We compare the obtained results withtheoretical expectations based on the bandwidth of participatingdisks. Mitra attains between 65% to 100% of the theoretical expectations.